机器人学研究室知识库

Intensity-based 2D/3D registration is a key technique using digitally reconstructed radiographs (DRRs) to register the preoperative volume to the patient setup during the operation. Although DRR-based method provides a high accuracy, the small capture range hinders its clinical use. In this paper, such problem was addressed by a robust and fast initialization method using a two-level scheme including automatic tracking-based initialization (Level I) and multiresolution estimation based on central-slice theorem and phase correlation (Level II). It provided almost optimal transformation parameters for intensity-based registration. Experiments using a public gold standard data set and a spinal phantom have been conducted. The mean target registration error (mTRE) was limited in the range from 2.12 mm to 22.57 mm after tracking-based initialization. The capture range based on level II only was 20.1 mm and the mTRE in this capture range was 2.92 +/- 2.21 mm. The intensity-based 2D/3D registration using proposed two-level initialization achieved the successful rate of 84.8% with the average error of 2.36 mm. The experimental results showed that the proposed method yielded the robust and fast initialization for intensity-based registration methods. In a similar way, it can be applied to other registration methods to enable a larger capture range and robust implementation.